Statistical mechanical algorithms (HAL, HALP, HALCO) were devised for evaluating amphipathic helical structures and more general structure- function relationships in proteins and peptides. These computational approaches are being used to define issues of structure and possible immunogenicity with respect to HLA antigens and to the envelope polyprotein of HIV. Also developed was a new joint prediction approach to protein secondary structure. The method combines neural network, information theory, sequence homology, hydrophobicity, loop potential, and amphipathicity criteria (including HAL). We have applied it to structure predictions on the human and bovine folate binding proteins, which are important in the cellular handling of folates and antitumor drugs such as methotrexate. Quantitative structure/activity relationship (QSAR) studies are being done on a set of 20 inhibitors of nucleoside transport into cells. In parallel, a molecular mechanics study is being done on analogues of the unusual nucleoside transport inhibitor dipyridamole. Our group has found that dipyridamole (and at least one other nucleoside transport inhibitor) potentiates the activity of azidothymidine against HIV-1 replication in cultured monocyte/macrophages and T-lymphocytes (see report # Z01 CB 08392-02 LMMB) . Animal studies are in progress and clinical trials are being planned, in parallel with continuing attempts to understand the mechanisms of action and molecular structure/function relationships.